Verification of dynamic properties of rule-based expert systems

A framework for the study of the dynamic properties of a class of rule-based expert systems is introduced. This is done by showing how the interconnection of a knowledge base with an inference mechanism can be viewed as a closed-loop control system. In this analogy the plant is the knowledge base and the controller is the inference mechanism. A discrete event system model is introduced and it is shown to be capable of modeling a wide class of knowledge bases and inference engines. The verification of reachability (achieving a chain of inferences) and cyclic behavior (e.g. circular reasoning) is discussed. A simple example is used to illustrate the approach and to point out the difficulties in performing formal verification